When LLMs prioritize the beginning and end of an input, and consequently perform poorly on middle sections.
literally put my prompt in the middle of the context and the ai just ignored it, lost in the middle is so real
+115
Think they got it wrong?
Add your own definition of Lost in the Middle